Cognitive control is important to everyday life, as we balance our desire to focus on a single task while putting other responsibilities aside and simultaneously resisting the urge to be distracted by more trivial tasks. Realistically, however, we rarely maintain focus on a single task, but instead switch between tasks. Despite this being a seemingly effortless process, switching incurs a cost, and research has suggested that at least some aspects our ability to switch between tasks changes with age. The present study
investigates how two components of task switching change with age.
In task switching paradigms, subjects traditionally complete two types of blocks. In single task (or pure) blocks, only a single task is relevant. In mixed task blocks, multiple tasks can be relevant throughout the duration of the block, though only a single task is relevant on a given trial, as specified by some sort of task cue. In such task switching paradigms, performance declines when switching between tasks is required, relative to when only a single task is performed. More specifically, this performance decline – the switch cost – is associated with executive control inherent to a multiple-task situation. The present study will address two costs in these sorts of task switching
paradigms: global switch costs and N-2 repetition costs (also referred to as backward inhibition).
Global switch costs are measured as the difference between mixed and pure blocks (e.g., Jersild, 1927; Rogers & Monsell, 1995; Spector & Biederman, 1976). These costs are hypothesized to measure the ability to update, manipulate, and maintain
multiple tasks in working memory (WM; Kray & Lindenberger, 2000; Mayr, 2001), as multiple tasks need to be available in the mixed block whereas only a single task is
relevant in pure blocks. According to Mayr (2001), global costs measure the ability to update “an internal control setting in the face of actual or potential interference” (p. 97) between tasks.
The second cost investigated herein is the N-2 repetition cost. This cost is thought to measure the application of inhibitory processes (backward inhibition) as a mechanism for disengaging from no-longer-relevant tasks at the point of a task switch. To measure task disengagement, Mayr and Keele (2000) used a task switching paradigms with three tasks (tasks A, B, and C). Critically, they compared trials in which task A was recently abandoned – trial sequence ABA – with trials in which task A was less recently
abandoned – trial sequence CBA. Mayr and Keele found worse performance on N-2 task repetition sequences (ABA), relative to N-2 task switch sequences (CBA), and this N-2 repetition cost has been taken as evidence for the use of backward inhibition for
disengaging from a no-longer-relevant task in order to switch to a new task (see also Gade & Koch, 2005, 2007; Schuch & Koch, 2003; Schneider & Verbruggen, 2008). That is, Mayr and Keele propose that task sets are inhibited once abandoned, therefore making it more difficult to switch back to a recently inhibited task (ABA), as opposed to a less recently inhibited task (CBA). Such hypothesized mechanisms, then, suggest a critical role for inhibition in task switching (Mayr, 2001), with some even suggesting that these backward inhibition mechanisms are more general, functioning also to clear the contents of WM when switching between WM items (Bao, Li, Chen, & Zhang, 2006).
Interestingly, both global and N-2 repetition costs are in some way associated with domains of executive control for which the evidence of decline in healthy aging has been the subject of controversy. In task switching, a large body of work has demonstrated
that older adults show exaggerated global switch costs relative to young adults, even when accounting for general slowing and practice effects (e.g., Kray & Lindenberger, 2000; Mayr, 2001; Mayr & Liebscher, 2001; Meiran, Gotler, & Perlman, 2001; Reimers & Maylor, 2005; Verhaeghen & Cerella, 2002; Wasylyshyn, Verhaeghen, & Sliwinski, 2011). Specifically, these authors have hypothesized that older adults have difficulty dealing with multiple task sets in working memory, specifically when the environment poses situations of task ambiguity. That is, in mixed blocks, tasks change relatively often, stimuli are complex and contain features relevant to multiple tasks, and all task responses are typically mapped to the same response modalities (e.g., the same two buttons). As a result, Mayr (2001) has hypothesized that older adults have difficulty selecting among relevant task sets in mixed blocks, especially under such conditions with difficult mappings between stimulus and response. In particular, Mayr hypothesized that age effects are exaggerated under conditions of task set ambiguity, such as when the stimulus is bivalent, has the potential to activate multiple task sets, and the cue does not explicitly indicate which task is relevant on a given trial (e.g., Rogers et al., 1998).
In contrast, a smaller number of studies have failed to find age differences in global costs (Brinley, 1965; Kray, Li, Lindenberger, 2002; Mayr & Kliegl, 2000; Salthouse, Fristoe, McGuthry, & Hambrick, 1998; Wheatley, Scialfa, Boot, Kramer, & Alexander, 2012; Verhaeghen & Basak, 2005; Verhaeghen & Hoyer, 2007). Kray et al. (2002) attribute cross-study differences to the degree of task uncertainty, with older adults having difficulty in selecting and instantiating a task under conditions of high competition between task sets. In their study, for example, on each trial, each of the four possible tasks were explicitly cued with a color coded letter that indicated the relevant
task, removing much task set selection uncertainty upon target presentation. Under these conditions, older adults did not demonstrate global shifting impairments. Thus, Kray et al. suggest that whether or not age differences in global shifting are found may be modulated by the degree of competition involved in choosing between tasks; age effects may only arise when distinct task sets contain overlapping elements and therefore elicit competition during task set selection, as task sets are not easily differentiated (see also, Mayr, 2001).
Similarly, while some have hypothesized that older adults show general inhibitory decline that has critical consequences for other areas of cognition (e.g., Hasher & Zacks, 1988), others have instead suggested that not all aspects of inhibition decline equally (see Chapter 4 for a more thorough discussion of this issue). In particular, the hypothesis of a general, all-encompassing inhibition deficit may be too broad, as inhibition may not be a unitary construct (e.g., Friedman & Miyake, 2004; Nigg, 2000). Additionally, to my knowledge, only one study has investigated whether older adults show impairments in backward inhibition (Mayr, 2001), as would be predicted by theories postulating general inhibitory decline. Using the backward inhibition paradigm described above, Mayr found little evidence for age-related impairments in inhibition, and the evidence he did find was not in the direction predicted by theories of inhibition deficits. That is, Mayr reasoned that if older adults have inhibition deficits, they should show reduced backward
inhibition effects: less (or inefficient) inhibition of task A in the N-2 repetition sequence (ABA) would result in easier reactivation of this task, when it again becomes relevant – and as a result, the difference between ABA and CBA conditions would be reduced or eliminated. Critically, Mayr found the opposite pattern of results; older adults showed
significantly larger backward inhibition effects than young adults when standard response times (RT) were used, though this effect was only marginal with log- transformed RTs (p = .07). In his study, Mayr compared the backward inhibition of young and old adults using a sample of twenty-four subjects per age group. While sizable for investigating simple cognitive effects, it may be the case that there was much
variability in the performance within such a small sample size, especially given that the older adults’ RT standard deviations were at least twice as large as those of the young adults. Before definitive conclusions about age-related performance on backward inhibition can be drawn, backward inhibition effects in young and old adults should be replicated. The present study investigates the global shifting and backward inhibition abilities of a large group of young and old adults.
Method
Participant, task, and data processing details are thoroughly described in the method chapter (Chapter 3). Additionally, background tasks assessing differences between the young and old adults samples are described and reported in Chapter 4. The present chapter discusses results from the three-task shifting task, which includes both pure and mixed task blocks; this task is reviewed briefly below. RTs reflect correct responses, excluding RTs on trials following errors. Analyses using RTs (in ms) excluded extreme outliers and outliers lying beyond 2.5 standard deviations of an individual
subject mean, for each condition. Log RTs reflect RTs that exclude extreme outliers only. Shifting Task
Subjects responded to the target based on the relevant cued task set. Targets were displayed one at a time in the center of the screen, and were preceded by cues that read
either “Number”, “Shape”, or “Size”. Subjects completed three pure blocks and three mixed blocks. In the pure blocks, subjects responded to either “Number, “Shape”, or “Size” throughout the duration of the block. In mixed blocks, subjects responded to any one of the three possible tasks, depending on the cue presented at the start of the trial, and the relevant task changed every trial. Subjects were first familiarized with the task in pure and mixed task practice blocks. Targets were selected pseudo-randomly with the
constraint that no exact stimulus repetitions were allowed. Additionally, the task sequence of mixed blocks was restricted such that a) all three tasks occurred equally often, b) there were neither direct task repetitions nor direct stimulus repetitions, c) each stimulus combination (size, shape, number; shape, size number; etc.) appeared equally often within a block, and d) there was an equal number of N-2 task switches (ABC) and N-2 task repetitions (ABA). The dependent variables were a) global switch costs, measured as the difference between mixed and pure blocks, and b) N-2 repetition costs (or backward inhibition), measured as the difference between N-2 repetition (ABA) and N-2 switch (CBA) trials (within the mixed block).
Results Analyses
In order to take baseline task performance into account, the effects of age on the shifting measures were investigated with a repeated measures ANOVA with age (young, old) as a between-subject factor and condition (for global costs, mixed vs. pure blocks; for backward inhibition, N-2 repetition vs. N-2 switch trial types) as a within-subjects factor. Tasks were separately analyzed using log RT data and errors. Log-transformed RTs were used to account for the slower processing speed of the older adults;
additionally, where relevant, processing speed was used as a covariate in error analyses. For all analyses, a main effect of condition was expected, replicating standard switching or repetition costs. Age x condition interactions that remained significant after controlling for processing speed were taken as evidence for age-related performance differences. Global Switch Costs
Figure 5.1 depicts performance in mixed and pure blocks, as a function of age, for both standard RT (top) and log-transformed (bottom) data. Subjects were slower and more error prone in mixed blocks (M = 1333 ms, 4%) relative to pure blocks (M = 578, 3%). This main effect of condition was significant for both log RT (F(1, 160) = 2006.88, MSE = 8.78, p < .001, η2 = .93) and errors (F(1, 160) = 33.69, MSE = 0.03, p < .001, η2 = .17). Additionally, the main effect of age was significant, as older adults were
significantly slower (M = 1412) than young adults (M = 921), F(1, 160) = 135.84, MSE = 2.25, p < .001, η2 = 46. However, the main effect of age was not significant in error rates (for both age groups, M = 4%), F(1, 160) = 1.24, MSE = 0.002, p = .27, η2 = .008. As shown in Figure 5.1 (top), older adults demonstrated larger RT switch costs than young adults; however, this interaction was not significant with log-transformed RTs, F(1, 160) = 0.33, MSE = 0.001, p = .57, η2 = .002. Additionally, the effect in errors only
approached significance, F(1, 160) = 3.35, MSE = 0.003, p = .07, η2 = .02, and it did not survive the inclusion of processing speed as a covariate, F(1, 159) = 0.12, MSE < 0.001, p = .73, η2 = .001.
Backward Inhibition Costs
Figure 5.2 depicts performance in N-2 switch (CBA) and N-2 repeat (ABA) trial types, as a function of age, for both standard RT (top) and log-transformed (bottom) data.
Subjects were slower and more error prone on N-2 repetition trials (M = 1431, 5% errors) relative to N-2 switch trials (M = 1364, 4%). This main effect of condition was
significant for both log RT (F(1, 160) = 173.31, MSE = 0.04, p < .001, η2 = .52) and errors (F(1, 160) = 7.81, MSE = 0.002, p = .006, η2 = .05). Older adults were
significantly slower (M = 1781) than the young adults (M = 1171), F(1, 160) = 80.68, MSE = 2.36, p < .001, η2 = .34. However, the two groups did not differ in error rates (for both, M = 4%), F(1, 160) = 0.03, MSE < 0.001, p = .85, η2 < .001). In RTs, older adults (M = 96 ms) showed larger backward inhibition costs than young adults (M = 49); however, this interaction did not survive log-transformation (F(1, 160) = .56, MSE < 0.001, p = .46, η2 = .003), nor was it significant in error rates (F(1, 160) = 1.63, MSE < 0.001, p = .20, η2 = .01).
General Discussion
The goal of the present study was to investigate age-related performance on two measures of task switching, global switch costs and N-2 repetition costs. While the effects of age on global shifting measures appear to be mediated by the ease with which task sets can be differentiated (e.g., Kray et al., 2002; Mayr, 2001 ), the effects of age on N-2 repetition costs have not been extensively investigated. Each measure is discussed below, in turn.
Global Switch Costs
Both age groups showed large global switch costs. Although older adults
demonstrated slower overall reaction times, they did not show exaggerated switch costs once general slowing was accounted for. In some ways, the present results were
global switch costs in older adults across many different shifting paradigms (e.g., Kray & Lindenberger, 2000; Mayr, 2001; see also Verhaeghen & Cerella, 2002, for a meta-
analysis that includes global costs as a function of age). These studies have suggested that older adults have difficulty in maintaining, manipulating, and selecting between multiple task sets, as needs to be done in mixed blocks, relative to pure blocks.
However, other work has suggested that age effects in global shifting may be reduced – or even eliminated – with reductions in task uncertainty. Kray et al. (2002), for example, did not find age differences in global costs using a cued shifting paradigm with verbal targets. In line with this, Mayr (2001) identified boundary conditions for age effects – when targets contained features of each possible task, and response mappings overlapped, substantial age effects were found. However, when these task set features did not overlap, age differences were minimal. As a result, Mayr suggested that older adults have to rely more strongly on task set updating processes, given they show worse
performance as a function of interference between task sets. In particular, Mayr raised the possibility that older adults have difficulties under conditions where task sets are not easily differentiated (see also Mayr & Liebscher, 2001), and that task set differentiation is more easily accomplished when there are close associations among task-relevant features such as possible responses, stimulus-responses mappings, etc.
That is, the notion that older adults may show deficits in measures of global shifting when binding processes play a critical role in differentiating among activated tasks in memory is consistent research from episodic and working memory domains (e.g., Chapter 4; Henkel, Johnson, & De Leonardis, 1998; Oberauer, 2005). In task switching, binding may function to create task sets, or distinct representations of what should be
done in the context of a given task – an attentional set that guides behavior and includes representations of what aspect of the target to focus on, possible responses, and stimuli- response mappings (e.g., Allport, Styles, & Hsieh, 1994; Rogers & Monsell, 1995). If binding among task features is weak (or fails), task sets may not be easily distinguishable, especially when task set features overlap. Thus, in global task switching settings, weak bindings may cause difficulty for older adults in distinguishing between relevant and irrelevant tasks, with failed binding mechanisms making it more difficult for older adults to form and/or differentiate between tasks, weak bindings increase the likelihood that the irrelevant task set will be inappropriately activated. Supporting the notion that older adults only seem to show global shifting impairments when task sets are not easily differentiated, Henkel et al. (1998) suggested older adults have the most difficulty in binding features in memory when distinct representations have some degree of overlap. Thus, task set binding mechanisms seem critical to task switching performance because shifting requires that task sets be kept apart (Mayr, 2001).
Like Kray et al. (2002), the present study failed to find exaggerated switch costs in older adults. However, the present study used explicit cues that may have provided environmental support, allowing older adults to more easily differentiate between task sets. Further supporting the notion that age effects may depend in part on the ability to differentiate between task sets, Kray, Eber, and Karbach (2008) found that task-relevant strategies (such as producing the name of the upcoming task) reduced (but did not eliminate) age effects. Thus, the design of the present study may have reduced the effect of any binding impairments in older adults. Interestingly, Mayr (2001) found age effects with an almost identical paradigm as used herein, even with explicit cues. Using a larger
group of subjects, the present study questions these results, suggesting that age-related impairments in global shifting are not always found – at least not when task sets can be relatively easily disambiguated.
Backward Inhibition
Replicating previous research, both young and old adults showed small but significant N-2 repetition costs. As with global switch costs, older adults were slower overall, but did not show exaggerated measures of backward inhibition once general slowing was accounted for. This replicates the findings of Mayr (2001), who found that older adults showed numerically larger backward inhibition effects, though such
differences were not significant once log-transformed. Using a larger sample, then, the present study supports the notion that older adults show no impairment in their ability to disengage from relevant task sets using inhibitory processes (see also Oberauer, 2005).
In one respect, the present results are surprising given findings and theoretical frameworks that propose that older adults have a deficit in inhibitory control processes (e.g., Hasher & Zacks, 1988). From this account, we might have expected older adults to show smaller backward inhibition effects, indicative of impaired inhibitory abilities. Instead, like Mayr (2001), older adults did not show exaggerated costs. Thus, the present results suggest that older adults do not show impairments in this measure of inhibition. As a result, older adults cannot be said to have a deficit to all inhibitory control
mechanisms (e.g., Hasher & Zacks, 1988), but may instead show selective deficits to some in only some of these mechanisms (e.g., Chapter 4). Additionally, these results suggest that backward inhibition is somehow distinct from those aspects of inhibitory control in which older adults do show age-related declines.
However, Anderson and Levy (2007) have suggested that the prediction of a relationship between backward inhibition and inhibitory control is quite difficult. As Anderson and Levy point out (and in line with the assumptions of Mayr & Keele, 2000), inhibition has both costs and benefits. The better one is at inhibition, the better one can inhibit previous tasks when switching to a new task. However, the better one is at inhibition, the more difficult it may be to retrieve a previously inhibited task set. In the case of comparing performance in the experimental condition (ABA; N-2 repetition) vs. the control condition (CBA; N-2 switch) in backward inhibition tasks, Mayr and
colleagues have focused on the difficulty in retrieving task A in the experimental condition, as this should be harder the better one is at inhibition. They have implicitly assumed that inhibiting task B should be of equivalent difficulty in the two conditions; as a result, the application of inhibition to task B should play no role in explaining
individual differences between these conditions. However, as Anderson and Levy point out, this is not the case. Given the prior inhibition of task A in the experimental (ABA) condition, this task set will have a lower activation strength relative to that for task B, than is the case for task A relative to task B in the control (CBA) condition. That is, the more recent inhibition of task A (in the experimental ABA condition) will result in a larger difference in activation levels between task B and task A, requiring more inhibition of task B. In contrast, the less recent inhibition of task A (in the control CBA condition) will result in a smaller difference in activation levels between task B and task A,
requiring less inhibition. The better one’s inhibition ability, the better one can suppress task B in the experimental condition with respect to task A, with this ability playing a